README.md

metadata

license: cc-by-4.0
language:
  - sr
pretty_name: Serbian WikiMedia dataset
size_categories:
  - 100M<n<1B
configs:
  - config_name: default
    data_files:
      - split: train
        path:
          - SrpKorNews.txt
task_categories:
  - text-generation

Highly curated, High-quality, Serbian news corpus

Each line represents document. Each Sentence in a document is delimited.

Dataset contains Serbian news source articles which were post-processed and corrected both automatically and manually. It contains around 468 million words.

from datasets import load_dataset
dataset = load_dataset("jerteh/SrpKorNews")

Preview:

print(dataset["train"][3088])
{'text': '"Srbija je dobro mesto za investicije i to bi trebalo iskoristiti, kako bi mladi ljudi imali dobru perspektivu", izjavio je tada Bufije na sastanku u Vizbadenu.'}

Citation:

@Inbook{CvRS2023LRS,
author="Krstev, Cvetana and Stankovi{\'{c}}, Ranka",
editor="Rehm, Georg and Way, Andy",
title="Language Report Serbian",
bookTitle="European Language Equality: A Strategic Agenda for Digital Language Equality ",
year="2023",
publisher="Springer International Publishing",
address="Cham",
pages="203--206",
isbn="978-3-031-28819-7",
doi="10.1007/978-3-031-28819-7_32"
}